GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
hmhm1229
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Run this model inference with full control and performance in your environment.
Model Details
Model Provider
hmhm1229
Model Tree
Qwen/Qwen2.5-VL-7B-Instruct
Input Modalities
Output Modalities
Supported Functionality
<|lcon|>python
import torchfrom conceptformer.retriever.modeling import ConceptFormerRetrievermodel = ConceptFormerRetriever.load("Qwen/Qwen2.5-VL-7B-Instruct",lora_name_or_path="hmhm1229/ConceptFormer-Qwen",pooling="eos",normalize=True,dtype=torch.bfloat16,device_map="auto",)model.eval()
The wrapper loads the adapter tokenizer before PEFT, resizes the base embedding table for
<|lcon|>, and merges the adapter. The base model is downloaded separately.
| Metric | InfoVQA | ChartQA | SlideVQA | TQA | OWID Charts | Wikimedia Maps | Average |
|---|---|---|---|---|---|---|---|
| Recall@10 | 93.03 | 98.67 | 90.88 | 72.20 | 99.24 | 76.04 | 88.33 |
| NDCG@10 | 79.23 | 95.79 | 82.41 | 41.30 | 95.39 | 61.69 | 75.97 |
Use this adapter with the ConceptFormer code repository. conceptformer_state.pt stores
the latent projection used by the training objective. Evaluation encodes images and
queries separately and retrieves only inside each dataset corpus.
Performance depends on document rendering, image resizing, and corpus version. This model is intended for retrieval research and should not be used as a factual QA system without downstream validation.